AI-BASED PERFORMANCE EVALUATION SYSTEM FOR GATKA PRACTITIONERS USING HUMAN POSE ESTIMATION AND SKELETON-BASED ACTION RECOGNITION

Abstract

Gatka is a traditional Sikh martial art that emphasizes coordinated body movements, weapon handling, defensive actions, and quick transitions. Conventional teaching and evaluation methods mainly depend on the experience and observation of instructors, making objective performance assessment difficult. This dissertation presents an Artificial Intelligence (AI)-based framework for the recognition and analysis of Gatka techniques using Human Pose Estimation and Machine Learning.

The proposed system employs the MediaPipe Pose model to extract human body keypoints from video frames. These keypoints are used to generate posture, joint-angle, and movement-based features, which are then classified using machine learning algorithms such as Support Vector Machine (SVM), Random Forest, and Neural Network. The framework recognizes important Gatka techniques, including basic stance, forward strike, defensive block, side movement, and circular movement.

Experimental results demonstrate that the proposed approach can accurately classify Gatka movements while providing real-time feedback for training and performance evaluation. The system also supports digital preservation of Gatka by creating a structured dataset and reducing dependence on manual assessment. This research highlights the potential of Artificial Intelligence and Computer Vision in modernizing traditional martial arts education. The proposed framework can be further extended for sports training, automated coaching, cultural heritage preservation, and intelligent motion analysis applications.